Recipe: Drug-target dossier in one call¶
A research team wants to retrieve what UniProt records about a protein. These annotations alone cannot establish drug-target tractability.
uniprot_target_dossier(accession) answers that in one call —
internally it fetches the UniProt entry plus the FASTA (two upstream
requests, not nine separate tool calls) and assembles a structured
nine-section report.
Request and result boundaries¶
Call uniprot_target_dossier(accession) using a verified UniProtKB
accession. Its nine views cover identity, function, derived sequence
chemistry, PDB/AlphaFold/InterPro links, ChEMBL/DrugBank references,
disease and variant annotations, GO evidence, and cross-references.
The tool does not run docking, evaluate biological efficacy, or independently infer druggability. A missing PDB or ChEMBL cross-reference on the UniProt entry does not establish that no such data exist elsewhere. Current numerical values require a live source retrieval; this repository does not maintain a separate drug-target atlas.
Why this composition matters¶
A drug-discovery team typically wants to answer:
- Which target-related records exist? ChEMBL/PDB cross-references guide research but cannot establish binding, druggability, or efficacy.
- Which disease annotations are present? UniProt disease records and MIM links are context, not independent clinical validation.
- Is it well-characterised? Cross-reference count + variant count + ECO-code diversity — rough completeness signal.
- What's the basic chemistry? MW / pI / hydrophobicity for buffer selection, expression-system choice — derived from the FASTA without an external tool.
The dossier puts all four on one page. The agent can then decide
which deeper tool calls are warranted (e.g. uniprot_get_variants
for the full variant table, uniprot_get_alphafold_confidence for
the per-band pLDDT, uniprot_get_evidence_summary for the
quality-of-evidence picture).
Pairing with other tools¶
| Question | Follow-up tool |
|---|---|
| "Show me the actual structures" | uniprot_resolve_pdb |
| "What model-wide pLDDT distribution is reported?" | uniprot_get_alphafold_confidence |
| "What variants are in ClinVar?" | uniprot_resolve_clinvar |
| "What papers cite this entry?" | uniprot_get_publications |
| "What do KEGG / OMA say about orthologs?" | uniprot_resolve_orthology |
Reproducibility¶
The dossier may use two UniProtKB responses: the entry JSON and a
FASTA sequence for derived chemistry. The main footer binds the entry;
when chemistry succeeds, its section separately reports the FASTA URL
and SHA-256. Machine-readable output carries chemistry_status and,
when applicable, chemistry_provenance.
Use uniprot_provenance_verify for each UniProt REST input separately,
with the original Accept header. A digest match establishes source
integrity, not independent scientific validation of derived values.
What this recipe is not¶
- Not a substitute for a target-validation pipeline. Real validation needs cell-based assays, in vivo models, off-target screens — none of which fit in a UniProt-gateway tool.
- Not a fitness-for-purpose claim. The dossier reports what UniProt knows; whether your specific drug-discovery program can translate that into a credible target hypothesis is a separate question.
The dossier is a source-attributed research overview, not a validated druggability or clinical interpretation.